Untangling the Sources of Congressional Partisanship: Ideology, Interests, and Opportunism

Transcription

1 Untangling the Sources of Congressional Partisanship: Ideology, Interests, and Opportunism Frances E. Lee Department of Government and Politics University of Maryland How much partisanship in Congress is caused by members political and strategic motives to support their parties, as opposed to their different ideological preferences? This project begins by recognizing that not all important policy conflict in Congress occurs on issues that divide liberals from conservatives. Capitalizing on this variation, an original dataset is created classifying roll-call votes on the basis of their ideological and issue content. The findings reveal that ideological disagreement only explains a portion of the partisanship observed in congressional voting. Ideological issues significantly increase partisan conflict in Congress. However, a great deal of partisan conflict occurs on issues that are not identifiably ideological, notably on proposals to reform basic rules of the electoral game and on procedural and parliamentary matters. Furthermore, presidential agenda-setting appears to be particularly important in provoking partisan conflict, as a president s fellow partisans attempt to produce policy successes on his agenda items and opposing partisans seek to deny him those successes, even on issues that would not be expected to divide liberals from conservatives. The study suggests that political interests and partisan opportunism exacerbate party conflict in Congress beyond what would be found if ideological differences were the sole source of disagreement. Paper prepared for the American Politics Workshop, Department of Government and Politics, University of Maryland, March 18,

2 Throughout its deliberations, the body was split into two fixed and violent parties.... In all questions, however unimportant in themselves, or unconnected with each other, the same names stand invariably contrasted on the opposite columns. Every unbiased observer may infer... that, unfortunately passion, not reason, must have presided over their decisions. When men exercise their reason coolly and freely on a variety of distinct questions, they invariably fall into different opinions on some of them. When they are governed by a passion, their opinions, if they are so to be called, will be the same. --James Madison, Federalist 50 Studies of public attitudes toward Congress have shown that the public perceives partisan conflict as unnecessary and as evidence that legislators do not have the public interest at heart (Hibbing and Theiss-Morse 1995; 2002). Like James Madison in Federalist 50, the American public tends to think of partisanship as bickering motivated by irrational hostility and partisan self-interest. If legislators reasoned through each public policy issue and drew their own individual conclusions, one would not expect to see two unified blocks of legislators standing opposed to one another across the wide range of issues before the Congress. Why should one party s members happen to agree with one another on so many different issues and disagree so frequently with members of the other party? From this perspective, the amount of partisan conflict in the contemporary Congress is evidence that members of Congress fail to think issues through on their own and simply behave as teams to promote their own partisan interests and undermine their opponents. Legislative scholars, however, are more willing than the public to assume a kind of good faith on the part of legislators. They tend to think that members of the two major parties in Congress vote differently because they hold different underlying preferences on the role or purpose of government. In its most rigorous form, the theory postulates that all legislators possess an ideal point on an abstract ideological spectrum ranging from liberal to conservative that constrains their positions across issues as they vote in Congress. If members ideal points lie 2

3 along a single dimension, then stable, patterned, and assuming members sort themselves into parties according to ideological preferences partisan voting coalitions in Congress would indeed result. 1 Legislative scholars generally agree that members ideological preferences are a principal cause of partisan behavior, regardless of whether they embrace a conditional party government model of legislative decisionmaking (Rohde 1991; Aldrich 1995) or a pure legislative majoritarianism (Krehbiel 1993; Brady and Volden 1998; 1998). From this point of view, legislative partisanship is not mere bickering, but principled disagreement about the role or scope of government power. Members of Congress, however, are not single-minded maximizers of their ideological preferences. They have other goals, including winning reelection and obtaining power in the chamber (Fenno 1973; Dodd 1977; Sinclair 1995), and cooperation with their fellow partisans affects their ability to achieve these goals. Even though legislative scholars are more willing than the public to grant that genuine differences of ideological preference drive legislative partisanship, they also recognize that members of Congress have strategic and electoral motives to divide along partisan lines, reasons separate from their level of intra-party agreement or interparty disagreement on ideology (Cox and McCubbins 1993, ; Gilmour 1995). As Aldrich (1995, 23) writes, The parties-in-government are more than mere coalitions of likeminded individuals. Indeed, even when partisans are not like-minded, they still have important political incentives to cooperate. These other goals are likely to create partisan behavior even in the absence of ideological agreement. 1 If ideological preferences cannot be ranged on a single dimension, formal theory would not predict stable voting coalitions. McKelvey s (1976) chaos theorem holds that when more than two decision makers make decisions on a policy that can be characterized using more than one dimension, there is no equilibrium, and any outcome could result, depending on which items appear on the agenda 3

4 The purpose of this study is to assess the importance of members political and strategic motives to support their parties, as opposed to their ideological preferences, in explaining congressional voting behavior. Recent scholarship on this topic has largely centered on whether partisan cooperation will produce policy outcomes that diverge from the median legislator s ideological position in favor of the majority party s median (Aldrich and Rohde 2000) or on whether party affiliation explains additional variation in individual members voting behavior after controlling for ideological preferences (Binder, Lawrence, and Maltzman. 1999; Snyder and Groseclose 2000). Instead, this study begins by noting that many of the issues considered in Congress do not speak to liberalism and conservatism at all, at least as those concepts are understood in the U.S. context, and cannot meaningfully be situated on an ideological spectrum (Matthews and Stimson 1975, 32-37). Ideology, therefore, cannot provide structure to congressional voting on those issues. If ideology alone drove partisanship, one would expect to find partisan behavior on ideological issues and nonpartisan behavior on non-ideological issues. If parties are held together by members common political interests as well as by their ideological preferences, however, one would expect to find more pronounced partisanship on those issues with identifiable ideological content than on issues without it, but partisan behavior regardless. If ideology has no effect on partisanship, one would expect to see the same amount of partisanship on average across the different types of issues without respect to their ideological content. To differentiate among these motivations for partisan behavior, this study employs an original dataset that classifies roll call votes on the basis of the ideological and political content of the issue. The findings reveal that ideological disagreement only explains a portion of the partisanship observed in congressional voting. Ideological issues significantly increase party 4

5 polarization and the likelihood of partisan conflict in Congress. However, substantial partisan conflict occurs on issues that are not identifiably ideological in content, notably on proposals to reform basic rules of the electoral game and on procedural and parliamentary matters. Furthermore, presidential agenda-setting appears to be particularly important in provoking partisan conflict, as a president s fellow partisans attempt to produce policy successes on his agenda items and opposing partisans seek to deny him those successes, even on types of issues that would not be expected in advance to divide liberals from conservatives. Taken together, the study suggests that partisan opportunism exacerbates conflict in Congress beyond what would be found if ideological differences were the sole source of disagreement. Ideology and the Organization of Political Conflict If ideology is to offer useful explanations of political behavior, it must be defined so that it can yield falsifiable empirical results. It is necessary, as Sartori (1969, 399) observes, to declare what ideology is not, i.e., to qualify the notion a contrario.... [I]f no such opposition is justified, then the notion of ideology loses much of its interest and has little explanatory value. If ideology is not defined, it becomes merely a tautological catch-all for behavioral patterns that are not explained in some other way, no less vague than the positing of culture as an explanation for political behavior, a theoretical approach often criticized by political scientists. With respect to congressional voting, it should be possible to identify at the outset, before observing voting patterns that emerge, which is the more conservative or the more liberal position on any given issue. If it is not possible, then ideology becomes nothing more than a post hoc summary description of voting patterns that affords no predictive or explanatory traction on political behavior. 5

6 Ideology may seem too complex and multifaceted for operational definition and systematic empirical study. Liberalism and conservatism have meant different things at different points in American history. 2 Liberals and conservatives often disagree among themselves about what true conservatism or true liberalism means. Moreover, liberal and conservative are clearly multidimensional terms in practice, even if the terminology itself posits a unidimensional scale from left to right. When one describes a contemporary Member of Congress as liberal, one suggests his position on scores of separate issues: he probably favors increases in the minimum wage; supports gun control measures; opposes tax cuts; interprets civil liberties broadly; supports affirmative action, abortion rights, and government policies that provide a social safety net. One would have to be naive to assume that all the positions that either liberals or conservatives hold emanate from a consistent set of philosophical principles. 3 As Converse (1964, ) writes, [I]t seems obvious that few belief systems of any range at all depend for their constraint upon logic.... What is important is that the elites familiar with the total shapes of these belief systems have experienced them as logically constrained clusters of ideas. Ideological issues for purposes of this study are identified as disputed issues included among the clusters of ideas about the proper role and purpose of government associated with contemporary liberalism and conservatism. Employing Converse s definition, an empirical study 2 Gerring (1998) analyzes the important ideological shifts in American party politics from In a surprising example of these shifts, Poole and Rosenthal (1997, 112) show that even an issue as central to the conservative world view today as opposition to the minimum wage was not mapped onto the left-right continuum in Congress prior to World War II. According to their analysis, the right party during the pre-war period did not oppose the minimum wage more than the left party did it took time for the issue to map onto the political spectrum as it does now. 3 Conservatives never tire of pointing out that liberals, who usually support broad interpretations of civil liberties, embrace a narrow construction of the Second Amendment or of the First Amendment in the case of hate speech. Liberals similarly enjoy pointing out apparent contradictions in conservative beliefs, often focusing on the tension between conservatives preference for narrow limits on governmental intervention in the marketplace, but broad governmental powers to regulate individual morals. 6

7 of the role of liberalism and conservatism in structuring conflict in Congress does not need to begin from first principles of political philosophy to derive definitions of ideology. The clusters of ideas associated with liberalism and conservatism need not possess strict logical or philosophical coherence in order to be studied empirically, as long as these ideas are stable, differentiated, and can be defined independently of the behavior one wants to explain with them. Most introductory textbooks of American government offer reasonably concise definitions of these clusters of ideas, and these definitions exhibit as much consistency across texts as do most other concepts in political science. Fiorina, Peterson and Voss (2005, 111, 113) provide one of the more comprehensive definitions: American liberalism supports an active government in the economic sphere but permits a high degree of autonomy when citizens make moral choices. Liberals endorse using political institutions to address widespread social inequalities, such as those associated with race or gender, but not to promote other forms of good behavior. They prefer more international cooperation to solve world problems... Conservatism, meanwhile, endorses less government regulation of economic matters but favors public policies that will shape the nation s culture in a moral direction rather than an amoral one.... They... endorse a strong military and an independent foreign policy focused on national security concerns. Although there is some variation in wording and emphasis across scholarly definitions of ideology, scholars largely agree that ideological disagreement between contemporary American liberals and conservatives focuses on the role of the government in: Redressing social and economic inequalities across groups and classes through government programs and progressive taxation; Regulating private economic activity for public purposes; Limiting individual freedoms in order to promote and preserve the traditional moral and social order; and, 7

8 With respect to foreign policy, increasing U.S. military power, authorizing its use, and exercising it unilaterally rather than through multinational institutions. These different clusters of ideas encompass a great many subjects, but by no means do they account for all the important topics on the congressional agenda. Many issues debated and voted on in Congress do not connect to these clusters of ideas, and positions on such issues cannot be meaningfully described as either liberal or conservative. First, the range of issues typically thought of as good government causes do not map onto the ideological spectrum. As Stokes observes, many of the issues that agitate our politics do not involve even a shriveled set of two alternatives of government action.... If we are to speak of a dimension at all, both parties and all voters [a]re located at a single point (Stokes 1963, 372). Liberals and conservatives do not hold different views on the need to prevent corruption in government. There is often substantial controversy between Republicans and Democrats in Congress over lobbying registration rules, voting technology and election recount rules, anti-voter fraud measures, congressional ethics, limits on gifts to members and their staffs, conflict of interest regulations, and campaign disclosure requirements, but it would not be appropriate to refer to the different positions members take in such disputes as ideological. Similarly, neither liberals nor conservatives support unbalanced budgets, per se. Both agree that government should responsibly balance revenues against expenditures, at least in the long run, even if they disagree on the level at which those balanced accounts should be set. There is no ideological disagreement on distaste for waste, fraud, and abuse in government. Liberals and conservatives often focus on different types of programs in rooting out these vices, but in principle no one favors them. Controversy over whether to spend more on the Superconducting Supercollider or the Tennessee Tombigbee Waterway takes the form of a dispute about cost- 8

9 benefit analysis and relative efficiencies, not over the role of government. Liberals and conservatives agree that government programs should be effective, well-administered, and welldesigned. Good government causes do not figure in textbook definitions of ideology because politicians do not hold opposing values on such matters, only differences of opinion about ways to achieve agreed-upon goals. Second, there are not consistent, identifiably liberal or conservative positions on the proper roles and procedures for the different institutions of American government. How broad is the president s power to reorganize the executive branch or to take military or other executive action without congressional approval? What are the appropriate rules governing majority and minority rights in congressional debate? What does advise and consent mean with respect to Senate approval of federal judges and high administrative appointments? Even federalism issues are similarly ambiguous. Liberalism in the U.S. has historically been associated with an expansive view of federal power, but when federal power runs afoul of liberal goals (e.g., when states adopt more restrictive environmental regulations, generous welfare or minimum wage policies, or more permissive medical marijuana or right to die policies than the federal government) liberals generally express a preference for states rights and flexibility, at least on those issues. The same pattern describes conservatives relationship to federalism. Although conservatism has historically been associated with a narrow view of federal power, conservatives generally abandon these concerns when a narrow definition will inhibit the achievement of their ideological goals (e.g., conservative initiatives to limit state courts roles in tort cases, federalize regulatory standards, or require states to adhere to strict welfare eligibility standards). Liberalism and conservatism do not offer clear guidance on institutional matters, even though these issues are perennially controversial in American politics. 9

10 Third, many federal domestic programs are neither regulatory nor designed to redress economic inequalities among groups or classes in American society, and therefore liberalism and conservatism do not present clear, differentiated positions on them: large-scale science projects; medical research and development; space exploration; transportation; flood control; irrigation; homeland security; military salaries, base construction and staffing; disaster relief; crop insurance; small business loans; and veterans benefits. In the nineteenth century there was intense ideological dispute over the proper role of the federal government with respect to internal improvements, but liberals and conservatives do not disagree in principle on such matters today. Some libertarian strands of conservatism may reject government s role in the provision of most public goods, but no one would posit this view as characteristic of mainstream American conservatism today. Liberals and conservatives may often dispute the amount of funds to devote to domestic programs, but Congress does more with respect to these programs than decide on the amount of funding for them. Liberals and conservatives do not embrace different views on whether non-regulatory, non-redistributive programs fall within government s proper authority and role, and therefore voting on such issues generally cannot be structured by members ideological preferences. Fourth, liberalism and conservatism do not speak to distributive policy, as defined by either Lowi (1964) or Wilson (1973). Distributive benefits are narrow in scope, and consequently their relationship to broad ideas about the proper role of government is highly ambiguous. Liberals and conservatives do not take clear differentiated positions on pork barrel projects or narrow tax breaks for particular industries, even though liberals are more favorable toward government spending generally, and conservatives tend to support tax reductions. In principle, of course, neither liberals nor conservatives support pork barrel spending or tax 10

11 loopholes. Such politics are quite clearly governed by constituency considerations, with members finding greater policy merit in the narrow programs and tax expenditures that benefit their constituents. Similarly, once a given level of federal funding for a program is established, liberalism and conservatism cannot in most cases guide members decisions on how the dollars should be divided geographically. 4 Such conflicts may be lengthy and intractable in Congress, but they cannot be meaningfully described as ideological. In sum, ideology should give order to many policy debates, but many important policy controversies fall outside the clusters of ideas associated with liberalism and conservatism in the U.S. What structure should we expect to find on those issues that ideology (as we know it) cannot organize? If ideological preferences explained all patterns in congressional voting, then votes on these issues should not exhibit any consistent pattern. If, even on those issues, we still find, as in Federalist 50, that the same names stand invariably contrasted on the opposite columns, and that those opposed lists of names tend to be members of different parties, then the more likely explanation is that the parties structured these conflicts. Non-Ideological Motives for Partisanship Ideological preferences produce patterned behavior in congressional voting, but they are not the only possible source of order. Members desire for reelection and career advancement gives them strategic and electoral incentives to cooperate with their parties and to find grounds on which to confront the opposing party on occasions when ideological preferences would not otherwise lead them to do so. Just as Krehbiel (1993) rightly observes that the incidence of 4 When the programs are designed to redress economic inequalities (Title I, Medicaid, Community Development Block Grants), conflict over distribution may be affected by members ideological preferences if Congress considers alternatives that would target aid to areas with greater poverty. In programs that do not have a redistributive purpose e.g., transportation, homeland security, research grants conflicts over distribution generally lack identifiable ideological content. 11

12 party voting alone offers no evidence that parties actually exercised any influence on members votes, one cannot conclude that simply because Republicans vote together on an issue while voting against Democrats (or vice versa) that different ideological preferences were the cause of the disagreement. It is always possible that the partisan pattern in voting had its source in members strategic or electoral goals, rather than in their diverging ideologies. Members incentives to cooperate with fellow partisans go far beyond the leadership pressure often referenced in the literature, which often confines meaningful party effects to leaders ability to use positive or negative incentives to induce members to bow to party discipline against their own policy preferences (Krehbiel 1993; Snyder and Groseclose 2000; McCarty, Poole, and Rosenthal 2001). Members of Congress have important individual strategic and electoral incentives to vote with their parties in the absence of whip pressure. First, members have an electoral interest in the collective image of their party. Members recognize that a substantial and perhaps growing proportion of the support they need to win elections derives from their association with one of the major political parties (Jacobson 2003, ). Members want to associate their party with popular policy outcomes and effective government and, equally valuable in a two-party system, to create the opposite associations with the opposing party. Members of Congress advance their reelection goals by cooperating with fellow partisans to improve their party s collective image or to tarnish the other party s image (Cox and McCubbins 1993, ). Partisan differentiation thus has electoral value to individual members of Congress because it gives voters reasons to support their party over candidates of the opposing party (Snyder 1994; Gilmour 1995; Wright and Schaffner 2002). Members... recognize a common stake in the image projected by the actions of the 12

13 congressional party, write Matthews and Stimson (1975, 97), and they feel a responsibility to strengthen the party position, whenever doing so does not conflict with their individual needs. Members want to deliver on an agenda that will improve the party s reputation, and this incentive is present even on issues that do not reflect ideological consensus within the party. Similarly, they will also seek to deny the opposing party high-profile policy successes that might improve its collective image. These considerations will produce partisan patterns in congressional voting, even when there are no ideological issues at stake. Moreover, members do not always put their policy preferences before their interests in reelection or career advancement in the chamber. Members were undoubtedly aware of their reelection and career interests in 2003 when considering a prescription drugs benefit for Medicare recipients. President Bush argued that passing the legislation would enable [Republicans] to say to seniors we kept our promise and thus improve the party s image on this issue (Earle 2003, 1). He gained the support of nearly all congressional Republicans for the largest expansion in government entitlement benefits since the Great Society. Meanwhile, congressional Democrats largely opposed the program, even though it clearly expanded government-provided social benefits relative to the status quo. As the leader of his party as well as the nation, the president plays a key role in defining the political agenda for both political parties. A presidential pronouncement on any policy issue raises the partisan stakes. Members of Congress suddenly have a political investment in the disposition of this issue, in addition to their individual policy preferences. Because of the issue s increased salience, presidential success on the issue is likely to enhance his party s reputation. Presidential success thus creates beneficial electoral spillovers for his party, giving his partisans 13

14 an interest in the president winning and members of the opposing party a stake in his losing, separate from whatever ideological preferences they might have. Second, members of Congress recognize that the rules of the electoral system e.g., congressional ethics, electioneering, and voter registration regulations can advantage or disadvantage their parties. Many of these issues are unconnected to ideological divisions in American politics. What would be the liberal position on Hatch Act regulations of federal employees involvement in political campaigns? What would be the conservative position on the rules governing voter registration of members of the Armed Forces? When votes are taken on such issues, Democrats are likely to perceive partisan benefits in relaxing Hatch Act restrictions because of the widespread perception that federal employees tend to be Democrats. Republicans are likely to support measures to increase voter registration in the Armed Forces because military personnel are seen as a Republican-leaning constituency. Regardless of ideology, members will act on common interests as fellow partisans in attempting to shape electoral and campaign rules to their own party s benefit. Third, members of Congress have motives to cooperate with their parties in order to increase their individual influence within the institution. Members power in the chamber is heavily dependent on their standing within their political parties. Political parties organize the Congress, and all members who obtain formal leadership positions do so through party auspices. With the weakened seniority system in Congress, obtaining desirable committee assignments or leadership positions has become even more dependent on being of assistance to the party. Ambitious members will calculate the likely effect of their votes on their career advancement in the chamber, providing further incentives to cooperate with their parties even in the absence of ideological motivations. Furthermore, members frequently take votes on parliamentary matters 14

15 (points of order, germaneness, appeals of the ruling of the chair) that affect the distribution of power over the congressional agenda between the majority and the minority parties. Members with ambitions to exercise power in the Congress cannot fail to take into account how their actions on parliamentary matters will affect their party s overall power stakes, either weakening or strengthening their influence (Binder 1997). 5 Taken together, these strategic and electoral motives will nurture bonds of trust and reciprocity between fellow partisans and sow skepticism and mistrust between members of opposing parties. Members do not have preferences on every issue that comes before the Congress. When they are uncertain about how to vote on an issue, members are likely to be swayed if they see an overwhelming majority of their own party voting in a particular direction (Matthews and Stimson 1975, 45). If they need additional information to make a voting decision, they do not turn at random to any member of the relevant committee of jurisdiction. They are far more likely to ask a member of the committee who is also a member of their party (Kingdon 1981, ). That fellow partisan is likely to give him many reasons to either support his party s position or to understand that the opposing party s position is flawed. These arguments may be based on common ideological preferences, but they may also be devoid of ideological content. Such arguments might turn on explicitly political reasons to assist one s party e.g., that a particular vote is important to the leadership, helpful to constituents, popular with key groups in the party s coalition or with the electorate at large. Nonideological policy arguments may simply catalogue the problems with the opposing party s plan, focusing on its flawed assumptions, poor policy design, duplication elsewhere, or waste and inefficiency. Members electoral and strategic goals are thus likely to result in patterns of trust and 5 This is not to imply that strengthening one s own party leadership s procedural control over the agenda will always be in individual members power interests; indeed, decentralized institutions may be preferable (Schickler 1997). However, members would always want to prevent the opposing party from centralizing power over the agenda. 15

16 communication that will greatly intensify partisan conflict beyond what would exist on the basis of the different ideological preferences of individuals. Data and Methodology Relationship to Previous Approaches In order to assess how congressional voting behavior is shaped by members electoral and strategic interests, as opposed to their ideological preferences, a dataset has been compiled classifying all roll-call votes in a series of Congresses on the basis of substantive ideological and issue content. Votes are classified based on criteria specified in advance of the observed voting behavior. This is not an approach that scholars have employed to study ideology in congressional voting. 6 Previous work on ideology and partisanship relies on either (1) interest group measures of ideology (such as the Americans for Democratic Action or the American Conservative Union ratings) or (2) variants of principal components analysis to map the distance between legislators in observed roll call voting behavior, used by the National Journal on subsets of roll call votes for their annual analyses of members voting behavior and by Poole and Rosenthal s (1991; 1997) NOMINATE procedure. Both types of measure are constructed post hoc, and neither is capable of distinguishing the circumstances when ideology causes partisan division from the occasions when purely political motivations would lead members to vote with their parties. Interest group measures of ideology cannot satisfactorily account for the role of ideology in congressional voting because they are created from small, often biased and idiosyncratic 6 This approach perhaps bears closest resemblance to Clausen s 1973 study of congressional decisionmaking that begins by classifying votes into five substantive policy areas. Clausen, however, altogether rejects ideological accounts of congressional voting as too gross a simplification of the decision-making apparatus of the typical congressman, so proud of his pragmatism and so scornful of ideological straitjackets (31). 16

17 subsets of votes. Interest groups have a political agenda in creating their scores; their motivation is not scholarly or analytical. As a result, scores include issues of concern to them whether or not scholars would consider them legitimate measures of legislators ideological preferences. 7 They select (and weight) votes in order to reward friendly legislators with favorable ratings. They routinely include votes in their ratings not just by virtue of the issues involved, but reversing the logical causal order and selecting on the dependent variable on the basis of whether the vote turned out in retrospect to have been close (Fowler 1982, ). Undoubtedly, some interest group ratings are more comprehensive and less biased than others, but political scientists should ideally develop their own measures of a concept as important as ideology. Measures of ideology that employ principal components analysis techniques are able to describe patterns in congressional votes after they have occurred, but they cannot discern whether the patterns observed were produced by members political interests, by their ideological commitments, or some combination thereof. The methodology maps the structure of congressional voting and locates where individual members fall within that structure; it cannot establish the cause of that structure when ideological, electoral, and strategic motives would all be expected to yield partisan voting alliances. Poole and Rosenthal (1997, 46) recognize the difficulty and stipulate that NOMINATE scores can be thought of as ranging from strong loyalty to one party... to weak loyalty to either party and to strong loyalty to the second 7 For example, the Americans for Democratic Action throughout the 1960s and 1970s routinely included legislators votes on cloture procedures, congressional committee reorganization, and other internal reform measures in rating members liberalism and conservatism (Zelizer 2004). The ADA probably thought that these reforms would enhance the chances that liberals would prevail on their ideological goals, but support for institutional reform should be kept analytically separate from a member s ideological commitments. After all, congressional reforms strengthening majority rule have been equally useful to the Republicans of the 1990s for achieving their conservative goals as they were to the Democrats of the 1970s for achieving their liberal goals, if not more so. 17

18 opposing party, and elsewhere (McCarty, Poole et al. 2001) that there is very substantial evidence that party affiliation has a strong influence on ideal points. An example from the National Journal ratings from 2003 illustrates one type of error that results from interpreting members place in the structure of congressional voting alliances as a measure of ideology: the vote on the expansion of Medicare to cover prescription drugs. Members divided along party lines in the vote on Medicare expansion, with most Republicans voting in favor of it and most Democrats against it, and so the analysis automatically classified members who voted in favor of it as taking the conservative position. The few hard line conservatives in the Congress who voted against their party leaders and their president actually found that this opposition resulted in scores that made them appear more liberal. Movement conservatives such as Rep. Jim DeMint (R-S.C.), Rep. Steve Chabot (R-Ohio), Sen. Trent Lott (R-Miss.) and Sen. Don Nickles (R-Okla.) all received less conservative scores as a direct result of their failure to support a dramatic expansion of a government entitlement program; meanwhile the Republican leaders in both chambers grounded out the conservative pole of the spectrum (Cohen 2004). Interpretive difficulties are not confined to ends against the middle voting (McRae 1970, 51), as in the 2003 Medicare prescription drugs vote. Votes can exhibit a partisan pattern, even if the issue at stake has no ideological content or even if the parties take positions that are the ideological reverse of what might be expected of the left and the right party. As an example of the latter, anti-lynching bills in the 1920s had a counterintuitive effect on members NOMINATE scores (Poole and Rosenthal 1997, 110). The Democrats, the left-leaning party on economic regulation at this time, also opposed anti-lynching and all other civil rights legislation. Because the pattern of votes on anti-lynching bills during this period followed party lines, with 18

19 the left party opposing civil rights legislation and the right party supporting it, any Republican who voted in favor of civil rights is treated for scaling purposes as having taken the conservative position and any Democrat who voted against civil rights scores as having taken the liberal position. 8 Principal components analyses of congressional voting are inherently ambiguous. One cannot say how much of the observable, largely one-dimensional structure of voting is the result of members party loyalty and how much is the result of ideological preferences. If congressional voting behavior were organized exclusively by members varying degrees of loyalty to their political parties, and ideology had no effect on voting whatsoever, the methodology would still yield estimates of members place in the structure of congressional voting, and it would still account for much of the variation in voting behavior. 9 A legislature in which two blocks of members led by opposing personalities did nothing other than wrangle over the spoils of patronage could be analyzed using this methodology. Voting in such a legislature would be unidimensional in structure, but the cause of that structure could only be party loyalty and not ideology, because the battles in the legislature were not about ideology. Statistical manipulations of raw voting data cannot address these fundamental problems of interpretation. Regardless of the reduction in error they afford in describing congressional voting, principal component analyses only provide a map of a congressional voting structure that is simultaneously and in unknown proportions organized by both ideology and party. 8 Civil rights bills track in and out of the first dimension over the course of American history. At times, voting patterns on these bills map onto the main line of cleavage between the two parties; at other times they divide the parties and are picked up by the second dimension (Poole and Rosenthal 1997, ). 9 It is worth noting that spatial voting analyses perform very poorly when the party system breaks down, most notably during the Era of Good Feelings of (Poole and Rosenthal 1997, 38-39). Similarly, voting in the nonpartisan Nebraska legislature exhibits little pattern, and the NOMINATE procedure produces little aggregate proportional reduction in error (Wright and Schaffner 2002, 373). 19

20 Coding Ideological Content Data collection begins from the premise that, if members are to use ideology as a decision shortcut, bills must have ideological content (Matthews and Stimson 1975, 34). The goal is to cull out all votes that deal with issues associated with liberalism and conservatism in contemporary U.S. politics. These include all votes dealing with the distribution of the tax burden, regulation of private economic activity for public purposes, government s share of the economy, social programs to redress inequalities, social issues (abortion, school prayer, affirmative action, crime, etc.), the amount of resources devoted to the U.S. military, the use of military force, and support for international treaties and multilateral institutions. The key coding criteria are whether the vote involves any of these broad categories of issues and whether it is possible to identify, using conventional definitions, which position is the more liberal or the more conservative. 10 The Senate is a better arena in which to examine how the content of issues under debate affects congressional voting behavior than the House because the rules of floor procedure in that body are far more permissive than in the House, where the majority party has full control over the floor and determines which issues members are able to vote on in the first place (Smith 1989; Oleszek 1996). The majority party leadership in the House is able to use special rules to prohibit the offering of amendments that have the potential to divide the majority party, inflating party unity in that chamber merely as an artifact of the systematic exclusion of divisive issues (Roberts and Smith 2003). The Senate s more open environment means that a wider range of issues are debated and voted on, thus providing a more complete picture of what types of issues unify and differentiate the two parties. 10 There are no inter-coder reliability issues, because one researcher (the author) coded all the votes in the study. However, I intend to train additional researchers to re-code random samples of cases in order to test the intersubjective validity of the coding. 20

21 Six consecutive Congresses were selected for study, the 102nd through the 107th Congresses, with an eye to producing every possible permutation of divided and unified party control: divided government with a Republican president (102nd), divided government with a Democratic president (104th-106th), unified Republican control (part of 107th) and unified Democratic control (103rd). No roll-call votes were excluded from the dataset. Each roll call vote is classified into one of six categories of ideological content based on the issues as posed in the text of the bill, resolution, or amendment and the issues raised in the debate in the Congressional Record. In most cases, the vote can be classified into one of the categories on the basis of legislative language. If it is not possible to make a determination from language alone, the debate in the Record on that issue is examined to see if members offer arguments or explanations that clarify that the issue falls into one of the categories. Most votes are taken on amendments, which are usually narrowly drafted, and so are relatively easy to code in the appropriate category. For votes on passage of legislation or amendments in the nature of a substitute, which often involve complex packages of policies, the coding rule used is whether the vote presents any issue that can be coded into one of the ideological categories. There is a category for ideological compromise, when a legislative package clearly involves gains for both liberals and conservatives (e.g., a minimum wage increase combined with tax cuts) and a category for votes involving more than one of these broad categories of issues. Because so many votes are taken on issues that divide economic conservatives from economic liberals, the economic policy category is subdivided into four additional categories for more refined analysis. Table 1 provides brief definitions of each category, along with examples of policy issues that fall into each. 21

22 [Table 1 here] For votes on nominations, appropriations, and trade agreements it is not possible to classify the vote into a category on the basis of legislative language alone, even though conflict often centers on grounds that fall into one of the specified categories. For these votes, therefore, I examine the debate published in the Congressional Record to determine if a senator raised issues that are identifiably ideological. If, for example, during a debate on an appropriations bill senators complain about its low or high levels of spending, or about a policy rider that falls into one of the above categories, the vote is classified into the relevant category. Similarly, nominees for high office are often controversial because of their record and views on economic, social or foreign policy; if the debate on the nomination reveals any senator expressing concern along these lines, then the nomination is coded into the appropriate category. When trade agreements are objected to because of other countries lack of environmental and labor protections, the dispute is coded as ideological in the economic policy category. Procedural and parliamentary votes (motions to table, motions to waive the budget act, motions to table, cloture, appeals of a ruling of the chair) are coded based on the underlying policy issue at stake. 11 If the underlying issue falls into one of the specified categories, then the procedural vote is coded as ideological. These votes are also flagged for separate analysis. Measuring Political Interests To measure members purely political interests in cooperating with their parties on particular congressional votes, four variables are included in the dataset. First, to measure the effect of presidential agenda-setting, a variable indicates whether the issue was mentioned in the 11 For example, Senate roll call vote 151 in the 107th Congress, 2nd session is a motion to waive the budget act to permit consideration of a permanent repeal of the estate tax. The vote is coded as both procedural and as ideological (in the distribution of the tax burden category). 22

23 presidential State of the Union Address immediately preceding the vote. Many presidential initiatives involve policies with clear ideological content: Patients Bill of Rights, Social Security privatization, Head Start, tax cuts, universal health care. However, presidents also champion issues that bear no relationship to the clusters of ideas that make up liberalism and conservatism in the U.S., issues on which liberals and conservatives do not have clear, differentiated positions: tests of student educational performance; missions to Mars; community policing; grant program consolidations; reinventing government and paperwork reduction; medical research increases; targeted tax credits for charitable contributions or education expenditures; homeland security; military pay and veterans benefits. When presidents mention an issue in the State of the Union Address, it gives it heightened salience, and presidential credibility is at stake. If fellow partisans support their president on such issues to provide their party with credit-claiming opportunities, while opposing partisans resist him in order to deny those opportunities for credit-claiming, then these issues will elicit heightened partisanship in congressional voting, even if they do not have ideological content. Second, a variable is included indicating whether the vote deals with basic rules of the political game. Some of these issues clearly involve the different understandings of the proper role and use of government power that divide liberals from conservatives: requiring broadcasters to donate free or reduced-cost television time to candidates, public financing of elections, and limits on amounts individuals and political action committees can donate to electoral campaigns. However, many of these issues have no connection to liberalism and conservatism: voting technology, congressional ethics, voter registration procedures, anti-voter fraud measures, Hatch Act, recount procedures, campaign disclosure requirements. Many of these issues have obvious or suspected consequences for the political parties fortunes, even if they lack ideological 23

24 content. If members act on their partisan interests while regulating the rules of the political game, these issues will spark increased levels of partisanship. Third, a variable is included indicating whether or not the vote was on a motion to table, to recommit to committee, or to waive the budget act. Framing votes in this way gives members of Congress distance on the underlying issue; it is a classic procedural ploy to avoid being recorded directly on politically sensitive policy issues (Oleszek 1996, 246). Senators are able to claim that they did not reach the merits of the matter; they merely decided that the Senate should not deal with it at this time. Handling issues as procedural motions drains some of the policy content out of them. If members policy preferences alone were the key to understanding congressional voting, then one might expect less partisanship (or ideological structuring) on these procedural motions, simply because their policy impact is much less clear. If, however, members of Congress find it helpful for their career advancement to ascertain the views of their party leaders and go along in the absence of contrary inclinations (Matthews and Stimson 1975, 95), then members are in a better position to support their party on these votes because their procedural nature gives members fewer reasons to have contrary inclinations. If members vote to support their parties as a default position unless strongly inclined otherwise, these votes should exhibit higher levels of partisanship. Fourth, a variable is included indicating whether the vote was on a parliamentary matter (a point of order, germaneness inquiry, cloture, appeal of the ruling of the chair, motion for attendance, motion to proceed or to adjourn, to suspend the rules of the Senate). Parliamentary votes are separated from the procedural matters above because, on the whole, they have even less policy content. They mainly deal with control of the Senate floor agenda and the distribution of power between the majority and the minority. If members support their party because they 24